Quantum speedup limits are no longer only a conjecture, Physical Superintelligence says: an unreleased Anthropic model produced a Lean-checked proof of the Aaronson-Ambainis conjecture on the third day of a run. Announced on November 6, 2026, the result means a quantum computer cannot save exponentially many looks at a random string on a yes-or-no question unless that string hides a rare pattern. For investors, it is a boundary on exponential advantage and a test of AI-found proofs.
Updated November 7, 2026
Key Takeaways
- An unreleased Anthropic model helped prove Aaronson-Ambainis in three days.
- Lean accepted the day-three proof and two later strengthenings.
- Exponential query speedups on random yes-or-no inputs need rare structure.
- The manuscript is public, but journal peer review is still ahead.
What do quantum speedup limits actually rule out?
Open since 2008, Scott Aaronson and Andris Ambainis asked when a quantum computer can be exponentially faster than an ordinary one on random inputs. If it held, an ordinary computer could match the quantum machine’s yes-probability, up to a small error, on all but a small fraction of random strings, with only polynomially more looks.
In PSI’s research announcement, the lab says the conjecture is now a theorem. An internal system, running instances of a non-public Anthropic model on PSI’s own infrastructure, had a complete proof during the third day. Lean, a proof assistant that checks every step, accepted it. The system then improved the result twice, and Lean accepted the final version.
The new bound is polynomial in the degree, where earlier general bounds decayed exponentially. Santage’s account puts the influence of some bit at least about 0.0025 times variance to the power 2.5, divided by degree to the power 6, for a non-constant bounded polynomial. That blocks an exponential gap in looks on unstructured yes-or-no problems. PSI’s close stays narrow: “Quantum computing still has plenty of mysteries, but now there is one fewer.”
Where can quantum computers still beat classical machines?
Dramatic speedups were never the random-input case. In 1994, Peter Shor found a quantum algorithm that factors large numbers far faster than any known method for ordinary computers, because the input hides a repeating pattern. On yes-or-no questions, PSI writes, a quantum computer cannot save exponentially many looks unless the string has a rare pattern.
| Setting | Query advantage | Status |
|---|---|---|
| Random yes-or-no inputs | At most polynomial | PSI, Nov. 6, 2026 |
| Fixed parallel rounds on random inputs | No exponential speedup | Stanford, August 2026 |
| Rare pattern in the input | Exponential, as in Shor | Known since 1994 |
Guy Blanc, Jordan Docter, Carmen Strassle and Li-Yang Tan, who proved the fixed-round case in August 2026, have been reading the full proof and are writing a shorter exposition. They wrote to PSI:
This is a huge result. It resolves a decades-old conjecture with important implications for the nature of problems that admit quantum speedups.
As analysis, exponential query speedups still require that rare pattern. They do not apply to arbitrary yes-or-no questions about unstructured bits. Quantum computers remain early, Business Insider notes, though they would still be a major upgrade for some tasks. Readers comparing timelines can start with our report on quantum risk to Bitcoin and Intel’s Classiq quantum software deal.
How credible is an AI-assisted math proof?
David Turturean built the research system, scoped the problem, ran it, prompted the first draft, and is responsible for the paper’s correctness. Advisor Jim Halverson audited the proof. Advisor Phiala Shanahan commented on the Lean formalization.
- WebProNews reports the manuscript and Lean formalization are public. Lean code is under Apache 2.0, while the paper remains copyrighted, and four theorems with no placeholders or added axioms were re-checked on two independent Lean kernels.
- Readers must still confirm that PrintedConjecture matches the conjecture as printed. The human-readable paper is not machine-checked.
- Santage notes there has been no journal peer review, and that a 2019 human-written proof was withdrawn within weeks after a fatal flaw in one lemma.
What should investors track next?
Business Insider reports that Terence Tao said last month that math’s “romantic era” was over, adding, “We are entering an era where exciting new things are possible, but some things are being lost.” As analysis, the quantum claim this result targets is exponential query advantage on random yes-or-no inputs. For AI as a research partner, price the verifier and the review queue, not only access to an unreleased model.
Frequently asked questions
What did Physical Superintelligence announce?
On November 6, 2026, PSI said a system using a non-public Anthropic model produced a Lean-checked proof of the Aaronson-Ambainis conjecture on day three.
Does this end exponential quantum advantage?
No. PSI says a quantum computer cannot save exponentially many looks on random yes-or-no questions unless the string has a rare pattern. Shor’s algorithm still needs structure.
Which Anthropic model produced the proof?
PSI has not named it, and Anthropic did not respond to Business Insider. The lab says only that instances of a non-public model were used.
Has the proof been peer reviewed?
Not in a journal, Santage notes. Lean accepted the proof, but readers must still confirm the formal statement matches the printed conjecture.